January 28, 2016
Ancient Chinese Game Mastered After Researchers Train Policy Network
Artificial intelligence isn’t so artificial anymore.
As technology continues to expand so too does the capabilities of what they can do when compared to the flesh and bone of the very people creating and programming them. Though it sounds like a science fiction concept, computers truly are getting smarter each day.
The latest example of this growing intelligence came out of Google’s DeepMind division which decided to program a computer to play the ancient Chinese game of Go. The marble game is centuries old and relies mainly of human intuition as players take turns to place black or white stones on a board, trying to capture the opponent’s stones or surround empty space to make points of territory. Computers, traditionally, have not fared well in the game.
Until, of course, Google researchers took on the challenge.
“We first trained the policy network on 30 million moves from games played by human experts, until it could predict the human move 57 percent of the time (the previous record before AlphaGo was 44 percent),” explained researchers David Silver and Demis Hassabis in a recent blog entry.
“But our goal is to beat the best human players, not just mimic them. To do this, AlphaGo learned to discover new strategies for itself, by playing thousands of games between its neural networks, and gradually improving them using a trial-and-error process known as reinforcement learning. This approach led to much better policy networks, so strong in fact that the raw neural network (immediately, without any tree search at all) can defeat state-of-the-art Go programs that build enormous search trees.”
The tests and trials, though, wouldn’t be complete without a human opponent. It wasn’t just any human opponent the Google team invited to compete, though, as reigning three-time European Go champion Fan Hui — an elite professional player who has devoted his life to Go since the age of 12 — was invited to Google’s London office for a challenge match.
“AlphaGo won by five games to zero — the first time a computer program has ever beaten a professional Go player<‘ the team explained.
W. Brice McVicar is a staff writer for SiteProNews.